The devil is in the data

Michelle Schonken, Investec

---

Many businesses are feasting on data, but starving for insights. According to Forrester Research, insights driven businesses are fundamentally different, transforming the information they have into value at every opportunity so that they can differentiate their products and services and offer a better customer experience.

For Canzius Pretorius, a partner in King Price’s data analytics team, running an insights-driven organisation demands that tech teams work side-by-side with the business. “If you just sit in the corner somewhere, you won’t have a proper grasp of the business’ problems and won’t be able to help them solve these challenges effectively.” The solutions are important, he says, but before you can even start talking about the tech needed to transform data into insights, you need to make sure that you have a relationship with business so that you can really understand what is happening on the ground. “If you don’t have this, none of the solutions you put in place will ever yield tangible results.”

Thinus van Rooyen, a senior data warehouse architect at DataGeek, agrees. If a business wants to produce the right type of data for analysis, the first step is to clearly define the questions that need to be answered by taking the time to understand the business’ pain points, as well as its goals and objectives.

There is a multitude of tooling out there that can be leveraged, but no single tool will address or meet all your needs.

Michelle Schonken, Investec

But if transforming data into business insights was easy, everyone would be doing it. With data volumes growing exponentially, it’s important to establish a strong data foundation so that you can manage everything effectively and make the right decisions at the right time, says Michelle Schonken, AI data lead at Investec. She believes that the data management pitfalls that modern businesses experience involve a lack of standards. If a business wants to use data to improve decision-making processes, the data has to be accurate, complete and consistent, she says. In addition, if you have misaligned data and business definitions and if you don’t have clear data ownership and accountability policies in place, you can’t understand the data context and won’t be able to use this information to deliver actionable insights. Everyone may be talking about automation and AI, but manual processes still exist across many organisations, which can result in inefficiencies and inaccuracies. And, if different departments operate in siloes and don’t look at the data holistically, it’s inevitable that you will have multiple versions of the same datasets, and multiple versions of the truth.

Pretorius echoes this sentiment. It’s essential to govern the process and agree on certain principles when you’re working with data, he says. As an insurer with many different business lines, consolidating data into a single view is a complex exercise because business processes across the different areas of the business are not exactly the same. “It can be very frustrating to sit at exco and have one person say that we did 100 sales and someone else say that we did 120 sales. More often than not, they’re both right, but the metrics they’re using to define sales might differ because of a lack of standards. This might not seem like a big issue, but it causes confusion and can easily derail the conversation because now everyone is less focused on sales and more focused on the accuracy of the data,” Pretorius says.

But when they do get things right, the results are clear. Pretorius says King Price recently implemented a new vehicle write-off model, which was aimed at saving costs for the company, as well as improving customer experience. What the model does, he says, is attempt to predict if a vehicle that was in an accident should be written off or not; the aim is to speed up the process. 

If you just sit in the corner somewhere, you won’t have a proper grasp of the business’ problems and won’t be able to help them solve these challenges effectively.

Canzius Pretorius, King Price

“Often, the client isn’t happy with the repairs on their car or there are a lot of hidden costs involved in the repair process – like car hire, for example,” he says. “The aim of this model is to predict – based on the accident conditions and the damages, as well as the potential salvage values we can recover if we were to sell the wreckage – if it’s smarter to write off the vehicle sooner.” These predications are based on internal data from the insurer’s claims processes, as well as industry data around salvage trends. Speeding up the process makes it simpler for the customer to move on from the accident and look for a new vehicle.

In your opinion, what data analysis tools are currently making the most impact?

Thinus van Rooyen, senior data warehouse architect, DataGeek:

One of the most commonly referenced reports on data analysis tools is Gartner’s Magic Quadrant for Analytics and Business Intelligence Platforms, which measures tools across 12 capabilities. I believe the capability that is going to become increasingly important is Natural Language Query. This capability has been part of the scoring criteria for a very long time, but, with the advances in Large Language Models (LLMs), it will undoubtedly have an increasing impact in the coming years. LLMs allow end-users to ask questions of the data directly rather than asking a data team to develop a dashboard or report, opening up a new mechanism for end-users to interact with the data.

Michelle Schonken, AI data lead, Investec:

No single tool will address or meet all your needs across the data analytics landscape, from visualising your data, to machine learning capabilities, data exploration, data management, data preparation, modelling and profiling. You may not believe it, but Excel is still one of the most widely used data analysis tools today. In my opinion, organisations need to explore and experiment with the tooling out there to determine and understand what will most effectively meet their needs. It’s also important to factor in who will be using these tools, both technical and non-technical users.

Canzius Pretorius, partner, King Price’s data analytics team:

Everybody wants to mention the latest, most fancy tech, but, if I’m honest, it’s probably Excel. The reason for this is that it’s very accessible. Everybody knows how to work with it and people derive a lot of value from it. As a techie, that’s not what I want to say to you, but I’m just being realistic. Excel is still a really powerful tool and many businesses tend to default to it to do the analysis they need. Power BI is also incredibly useful.

The mining industry faces many challenges when it comes to managing, protecting and storing data, say Hemant Harie, Group CTO, Gabsten Technologies, and Iniel Dreyer, MD, Data Management Professionals South Africa. According to EY, cyber threats are evolving and escalating for businesses within the mining, metals and asset-intensive industries. Just ask Rio Tinto and Anglo American. In March 2023, Rio Tinto reported that the personal data of some past and present staff had been stolen as part of a cyber attack. Similarly, in December last year, subscribers to Anglo American’s email newsletter were sent a message from the company telling them to “GO F*** YOURSELF” after the company’s email distribution channels were compromised. While hacks like this are a major problem, poor data management also has potential to affect mine safety. Should something happen to the critical data systems that measure the temperature or gas in mine shafts, for example, they won’t function as they should and could put lives at risk.

At its core, master data management (MDM) is about creating a single, verified and trusted view of your data. Whether it’s data about a customer, about sales performance or about a problem the business faces, the goal is to understand the subject as a whole. MDM provides consistency, accuracy, stewardship, semantic coherence and accountability across a company’s data assets, says Mike Saunders, CEO, Digitlab. MDM prevents errors, improves data quality and supports business initiatives and decisions. It’s crucial in larger and more complex organisations, where data is often siloed across different departments, and eliminates fragmentation and inconsistent data.